Impact and Analysis of Weather Events
CSV Pitchers
December 5, 2017
Storm Event Database
About Dataset
The National Centers for Environmental Information (NCEI) regularly receives Storm Data from the National Weather Service (NWS). This data set consists of state wise occurrence of Event type in USA like Tornado, Thunderstorm Wind and Hail, Marine strong wind, Flash flood, Heavy rain, Heavy snow, Funnel Cloud, Extensive Heat and also contains data regarding locations, fatalities, injuries, damage, narratives and any other event specific information which can be used for information and analysis by business sectors, insurance companies, hazard mitigation, policy makers etc. (NCDC, 2017)
Data for Analysis
We are using the data of 2016 to analyze the effect of event types on economy (property damage and crop damage) and casualties (injuries and deaths). Furthermore, We have used data of 2011-2016 to see the trends of top events we have found in 2016.Overall there are 51 attributes/columns in each data file.
License
The information on NOAA web page is in the public domain unless specifically annotated otherwise (copyright may be held elsewhere) and may therefore be used freely by the public.
Additional information disclaimer and copyright notice can be found in Disclaimer and Copyright Notice .
Metadata : Information available to understand and interprete data
Storm events database contains the records with information on
- Occurrence of Storms and other significant events which is likely to cause more damage, death and injuries.
- Other rare, unusual weather phenomenon like snow flurries
The data is entered by NOAA’s National Weather Service(NWS). Overall 48 event types are recorded as metnioned in NWS Directive 10-1605.
51 Columns/datafields are there in data which can be defined as shown in Storm Data Export Format
Issues Encountered with data
- Columns having similar properties are there such as BEGIN_YEARMONTH, BEGIN_DAY, BEGIN_TIME and BEGIN_DATE_TIME. Similar case is found for END DATE as well.
- When looking into 2016 data, 11690 out of 55959 have blank values in DAMAGE_PROPERTY though it also have values 0.00 K in other rows.
- 11355 out of 55959 have blank values in DAMAGE_CROPS in 2016 data. Like in DAMAGE_PROPERTY, this field also contains value 0.00K.
- Negligible row i.e. 45 out of 55959 data in 2016 have source UNKNOWN.
Remediation steps
*We are using different set of columns for each of the research question so that our analysis is oriented towards a particular audience and our result can be implemented easily . For our intitial analysis we are removing ‘none’ value because R automatically excludes all cases in which any of the inputs are missing; this can limit the amount of information available in the analysis.
*We are dropping the columns which have redundant data such as ‘BEGIN_YEARMONTH’, ‘END_YEARMONTH’, ‘state_fips’ as these values are combination of columns present in the data set and for the analysis we can keep the data granular so that we can represent the result in more granular way.
*We are also removing the columns ‘EPISODE_NARRATIVE’, ‘event_narrative’ and ‘DATA_SOURCE’ as these are nominal data and it will hard to analyize them in R, but we will be using them seperately as remarks. We have checked for the duplicate values and removed the duplicate values for each question.
Data Cleaning & Analysis Steps
Loading the data
We have used the necessary functions and loaded the data from 2011-2016 into the respective dataframe as:
dataset2016 <- read.csv('StormEvents_details-ftp_v1.0_d2016_c20170918.csv',header = TRUE,stringsAsFactors = F, na.strings = c("",NULL,'NA'))
dataset2015 <- read.csv('StormEvents_details-ftp_v1.0_d2015_c20170918.csv',header = TRUE,stringsAsFactors = F,na.strings = c("",NULL,'NA'))
dataset2014 <- read.csv('StormEvents_details-ftp_v1.0_d2014_c20170718.csv',header = TRUE,stringsAsFactors = F,na.strings = c("",NULL,'NA'))
dataset2013 <- read.csv('StormEvents_details-ftp_v1.0_d2013_c20170519.csv',header = TRUE,stringsAsFactors = F,na.strings = c("",NULL,'NA'))
dataset2012 <- read.csv('StormEvents_details-ftp_v1.0_d2012_c20170519.csv',header = TRUE,stringsAsFactors = F,na.strings = c("",NULL,'NA'))
dataset2011 <- read.csv('StormEvents_details-ftp_v1.0_d2011_c20170519.csv',header = TRUE,stringsAsFactors = F,na.strings = c("",NULL,'NA'))These data are then combined into one dataframe as :
dataAll <- rbind(dataset2016,dataset2015,dataset2014,dataset2013,dataset2012,dataset2011)Internal structure of data has been checked
str(dataAll)## 'data.frame': 376788 obs. of 51 variables:
## $ BEGIN_YEARMONTH : int 201607 201607 201607 201607 201607 201607 201612 201612 201612 201602 ...
## $ BEGIN_DAY : int 15 15 16 8 8 8 4 4 4 25 ...
## $ BEGIN_TIME : int 1715 1725 1246 1755 1810 1910 300 400 400 1157 ...
## $ END_YEARMONTH : int 201607 201607 201607 201607 201607 201607 201612 201612 201612 201602 ...
## $ END_DAY : int 15 15 16 8 8 8 5 5 5 25 ...
## $ END_TIME : int 1715 1725 1246 1755 1810 1910 600 500 500 1841 ...
## $ EPISODE_ID : int 108769 108769 108812 105872 105872 105872 111575 111575 111575 103607 ...
## $ EVENT_ID : int 651823 651825 651828 635287 635296 635301 665685 665686 665687 619257 ...
## $ STATE : chr "SOUTH CAROLINA" "SOUTH CAROLINA" "NORTH CAROLINA" "TENNESSEE" ...
## $ STATE_FIPS : int 45 45 37 47 47 47 56 56 56 24 ...
## $ YEAR : int 2016 2016 2016 2016 2016 2016 2016 2016 2016 2016 ...
## $ MONTH_NAME : chr "July" "July" "July" "July" ...
## $ EVENT_TYPE : chr "Heavy Rain" "Thunderstorm Wind" "Thunderstorm Wind" "Thunderstorm Wind" ...
## $ CZ_TYPE : chr "C" "C" "C" "C" ...
## $ CZ_FIPS : int 91 91 111 105 155 7 1 12 13 33 ...
## $ CZ_NAME : chr "YORK" "YORK" "MCDOWELL" "LOUDON" ...
## $ WFO : chr "GSP" "GSP" "GSP" "MRX" ...
## $ BEGIN_DATE_TIME : chr "15-JUL-16 17:15:00" "15-JUL-16 17:25:00" "16-JUL-16 12:46:00" "08-JUL-16 17:55:00" ...
## $ CZ_TIMEZONE : chr "EST-5" "EST-5" "EST-5" "EST-5" ...
## $ END_DATE_TIME : chr "15-JUL-16 17:15:00" "15-JUL-16 17:25:00" "16-JUL-16 12:46:00" "08-JUL-16 17:55:00" ...
## $ INJURIES_DIRECT : int 0 0 0 0 0 0 0 0 0 0 ...
## $ INJURIES_INDIRECT : int 0 0 0 0 0 0 0 0 0 0 ...
## $ DEATHS_DIRECT : int 0 0 0 0 0 0 0 0 0 0 ...
## $ DEATHS_INDIRECT : int 0 0 0 0 0 0 0 0 0 0 ...
## $ DAMAGE_PROPERTY : chr "2.00K" "0.00K" "0.00K" NA ...
## $ DAMAGE_CROPS : chr "0.00K" "0.00K" "0.00K" NA ...
## $ SOURCE : chr "Social Media" "NWS Employee" "Broadcast Media" "Public" ...
## $ MAGNITUDE : num NA 50 50 50 60 50 NA NA NA NA ...
## $ MAGNITUDE_TYPE : chr NA "EG" "EG" "EG" ...
## $ FLOOD_CAUSE : chr NA NA NA NA ...
## $ CATEGORY : int NA NA NA NA NA NA NA NA NA NA ...
## $ TOR_F_SCALE : chr NA NA NA NA ...
## $ TOR_LENGTH : num NA NA NA NA NA NA NA NA NA NA ...
## $ TOR_WIDTH : num NA NA NA NA NA NA NA NA NA NA ...
## $ TOR_OTHER_WFO : chr NA NA NA NA ...
## $ TOR_OTHER_CZ_STATE: chr NA NA NA NA ...
## $ TOR_OTHER_CZ_FIPS : int NA NA NA NA NA NA NA NA NA NA ...
## $ TOR_OTHER_CZ_NAME : chr NA NA NA NA ...
## $ BEGIN_RANGE : int 1 1 2 1 1 1 NA NA NA 0 ...
## $ BEGIN_AZIMUTH : chr "N" "S" "ENE" "W" ...
## $ BEGIN_LOCATION : chr "BOYD HILL" "FT MILL" "OLD FT" "JENA" ...
## $ END_RANGE : int 1 1 2 1 1 1 NA NA NA 0 ...
## $ END_AZIMUTH : chr "N" "S" "ENE" "W" ...
## $ END_LOCATION : chr "BOYD HILL" "FT MILL" "OLD FT" "JENA" ...
## $ BEGIN_LAT : num 34.9 35 35.6 35.6 35.9 ...
## $ BEGIN_LON : num -81 -80.9 -82.1 -84.2 -83.8 ...
## $ END_LAT : num 34.9 35 35.6 35.6 35.9 ...
## $ END_LON : num -81 -80.9 -82.1 -84.2 -83.8 ...
## $ EPISODE_NARRATIVE : chr "Scattered thunderstorms developed across the Upstate during the late afternoon and evening. Some of the storms "| __truncated__ "Scattered thunderstorms developed across the Upstate during the late afternoon and evening. Some of the storms "| __truncated__ "Scattered thunderstorms developed near the Blue Ridge during the afternoon and moved slowly east some of the st"| __truncated__ "The combination of summer heating and high low level moisture content provided the necessary conditions to crea"| __truncated__ ...
## $ EVENT_NARRATIVE : chr "One to two feet of deep standing water developed on a street on the Winthrop University campus after more than "| __truncated__ "NWS Columbia relayed a report of trees blown down along Tom Hall St." "Media reported two trees blown down along I-40 in the Old Fort area." "Numerous trees were reported down in the Greenback area." ...
## $ DATA_SOURCE : chr "CSV" "CSV" "CSV" "CSV" ...Duplicates have been checked
dupAll<- unique(dataAll)
nrow(dataAll)## [1] 376788nrow(dupAll)## [1] 376788NA is replaced with 0
dataAll_na <- dataAll
dataAll[is.na(dataAll)] <- 0Values of amount in damage property and damage crops containing M in data is multiplied by 1000 to turn it into K and “K” and “M are then removed from data
dataall_update <- dataAll
index1 <- (substr(dataall_update$DAMAGE_PROPERTY,nchar(dataall_update$DAMAGE_PROPERTY),nchar(dataall_update$DAMAGE_PROPERTY)))=="M"
index2 <- (substr(dataall_update$DAMAGE_CROPS,nchar(dataall_update$DAMAGE_CROPS),nchar(dataall_update$DAMAGE_CROPS)))=="M"
index3 <- (substr(dataall_update$DAMAGE_PROPERTY,nchar(dataall_update$DAMAGE_PROPERTY),nchar(dataall_update$DAMAGE_PROPERTY)))=="K"
index4 <- (substr(dataall_update$DAMAGE_CROPS,nchar(dataall_update$DAMAGE_CROPS),nchar(dataall_update$DAMAGE_CROPS)))=="K"
dataall_update$DAMAGE_PROPERTY[index1] <-
as.numeric(substr(dataall_update$DAMAGE_PROPERTY[index1],1,nchar(dataall_update$DAMAGE_PROPERTY[index1])-1))*1000
dataall_update$DAMAGE_CROPS[index2] <-
as.numeric(substr(dataall_update$DAMAGE_CROPS[index2],1,nchar(dataall_update$DAMAGE_CROPS[index2])-1))*1000
dataall_update$DAMAGE_PROPERTY[index3] <-
as.numeric(substr(dataall_update$DAMAGE_PROPERTY[index3],1,nchar(dataall_update$DAMAGE_PROPERTY[index3])-1))
dataall_update$DAMAGE_CROPS[index4] <-
as.numeric(substr(dataall_update$DAMAGE_CROPS[index4],1,nchar(dataall_update$DAMAGE_CROPS[index4])-1))Fields Damage property and Damage crops have to be changed into numeric datatypto allow any further analysis
dataall_update$DAMAGE_PROPERTY <- as.numeric(as.character(dataall_update$DAMAGE_PROPERTY))
dataall_update$DAMAGE_CROPS <- as.numeric(as.character(dataall_update$DAMAGE_CROPS))Replacing NA with 0
dataall_update_na <- dataall_update
dataall_update[is.na(dataall_update)] <- 0Relevant Statistics are being checked, in this case,summary is taken.
We have to mainly focus on damage property, damage crops, direct injuries, indirect injuries, direct death and indirect deaths.
summary(dataall_update)## BEGIN_YEARMONTH BEGIN_DAY BEGIN_TIME END_YEARMONTH
## Min. :201101 Min. : 1.00 Min. : 0 Min. :201101
## 1st Qu.:201204 1st Qu.: 6.00 1st Qu.: 700 1st Qu.:201204
## Median :201308 Median :15.00 Median :1450 Median :201308
## Mean :201338 Mean :14.85 Mean :1258 Mean :201338
## 3rd Qu.:201505 3rd Qu.:23.00 3rd Qu.:1800 3rd Qu.:201505
## Max. :201612 Max. :31.00 Max. :2359 Max. :201612
## END_DAY END_TIME EPISODE_ID EVENT_ID
## Min. : 1.00 Min. : 0 Min. : 45914 Min. :266353
## 1st Qu.: 9.00 1st Qu.:1112 1st Qu.: 62087 1st Qu.:372418
## Median :18.00 Median :1609 Median : 78298 Median :472047
## Mean :17.04 Mean :1485 Mean : 78525 Mean :472064
## 3rd Qu.:25.00 3rd Qu.:1919 3rd Qu.: 94922 3rd Qu.:571798
## Max. :31.00 Max. :2359 Max. :117998 Max. :709335
## STATE STATE_FIPS YEAR MONTH_NAME
## Length:376788 Min. : 1.00 Min. :2011 Length:376788
## Class :character 1st Qu.:19.00 1st Qu.:2012 Class :character
## Mode :character Median :31.00 Median :2013 Mode :character
## Mean :32.69 Mean :2013
## 3rd Qu.:46.00 3rd Qu.:2015
## Max. :99.00 Max. :2016
## EVENT_TYPE CZ_TYPE CZ_FIPS CZ_NAME
## Length:376788 Length:376788 Min. : 1.00 Length:376788
## Class :character Class :character 1st Qu.: 25.00 Class :character
## Mode :character Mode :character Median : 63.00 Mode :character
## Mean : 99.55
## 3rd Qu.:117.00
## Max. :876.00
## WFO BEGIN_DATE_TIME CZ_TIMEZONE
## Length:376788 Length:376788 Length:376788
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
##
##
##
## END_DATE_TIME INJURIES_DIRECT INJURIES_INDIRECT
## Length:376788 Min. :0.00e+00 Min. : 0.00000
## Class :character 1st Qu.:0.00e+00 1st Qu.: 0.00000
## Mode :character Median :0.00e+00 Median : 0.00000
## Mean :5.29e-02 Mean : 0.01306
## 3rd Qu.:0.00e+00 3rd Qu.: 0.00000
## Max. :1.15e+03 Max. :300.00000
## DEATHS_DIRECT DEATHS_INDIRECT DAMAGE_PROPERTY
## Min. : 0.0000 Min. : 0.000000 Min. : 0.0
## 1st Qu.: 0.0000 1st Qu.: 0.000000 1st Qu.: 0.0
## Median : 0.0000 Median : 0.000000 Median : 0.0
## Mean : 0.0092 Mean : 0.002896 Mean : 131.5
## 3rd Qu.: 0.0000 3rd Qu.: 0.000000 3rd Qu.: 0.0
## Max. :158.0000 Max. :13.000000 Max. :910000.0
## DAMAGE_CROPS SOURCE MAGNITUDE
## Min. : 0.00 Length:376788 Min. : 0.00
## 1st Qu.: 0.00 Class :character 1st Qu.: 0.00
## Median : 0.00 Mode :character Median : 0.88
## Mean : 36.99 Mean : 18.08
## 3rd Qu.: 0.00 3rd Qu.: 50.00
## Max. :250000.00 Max. :174.00
## MAGNITUDE_TYPE FLOOD_CAUSE CATEGORY
## Length:376788 Length:376788 Min. :0.000000
## Class :character Class :character 1st Qu.:0.000000
## Mode :character Mode :character Median :0.000000
## Mean :0.000318
## 3rd Qu.:0.000000
## Max. :5.000000
## TOR_F_SCALE TOR_LENGTH TOR_WIDTH
## Length:376788 Min. : 0.00000 Min. : 0.000
## Class :character 1st Qu.: 0.00000 1st Qu.: 0.000
## Mode :character Median : 0.00000 Median : 0.000
## Mean : 0.06504 Mean : 3.854
## 3rd Qu.: 0.00000 3rd Qu.: 0.000
## Max. :100.00000 Max. :4576.000
## TOR_OTHER_WFO TOR_OTHER_CZ_STATE TOR_OTHER_CZ_FIPS
## Length:376788 Length:376788 Min. : 0.0000
## Class :character Class :character 1st Qu.: 0.0000
## Mode :character Mode :character Median : 0.0000
## Mean : 0.2756
## 3rd Qu.: 0.0000
## Max. :510.0000
## TOR_OTHER_CZ_NAME BEGIN_RANGE BEGIN_AZIMUTH BEGIN_LOCATION
## Length:376788 Min. : 0.0 Length:376788 Length:376788
## Class :character 1st Qu.: 0.0 Class :character Class :character
## Mode :character Median : 0.0 Mode :character Mode :character
## Mean : 1.5
## 3rd Qu.: 2.0
## Max. :185.0
## END_RANGE END_AZIMUTH END_LOCATION BEGIN_LAT
## Min. : 0.000 Length:376788 Length:376788 Min. :-14.40
## 1st Qu.: 0.000 Class :character Class :character 1st Qu.: 0.00
## Median : 0.000 Mode :character Mode :character Median : 33.46
## Mean : 1.521 Mean : 23.47
## 3rd Qu.: 2.000 3rd Qu.: 39.30
## Max. :185.000 Max. : 70.50
## BEGIN_LON END_LAT END_LON EPISODE_NARRATIVE
## Min. :-170.89 Min. :-14.39 Min. :-170.83 Length:376788
## 1st Qu.: -92.85 1st Qu.: 0.00 1st Qu.: -92.85 Class :character
## Median : -80.52 Median : 33.46 Median : -80.51 Mode :character
## Mean : -55.97 Mean : 23.46 Mean : -55.96
## 3rd Qu.: 0.00 3rd Qu.: 39.29 3rd Qu.: 0.00
## Max. : 171.37 Max. : 70.26 Max. : 171.37
## EVENT_NARRATIVE DATA_SOURCE
## Length:376788 Length:376788
## Class :character Class :character
## Mode :character Mode :character
##
##
## Now we will start working on data and transforming it accordingly to answer the research questions for analysis. #### 1. Over the year 2016, what is frequency of occurrence of event type on the basis of location?
For this analysis 2016 data has been used to determine the number of even types that occurred on the basis of location. In the data state, longitudes and latitudes are available for the information on location.
library(knitr)
library(ggplot2)
# making subset to extract state and event type
Dataset_new <- dataset2016[,c(9,13)]
# Added new column Named region for States-
Dataset_new$Region[Dataset_new$STATE %in% c("CONNECTICUT","MAINE","MASSACHUSETTS", "NEW HAMPSHIRE","RHODE ISLAND", "VERMONT", "NEW JERSEY", "NEW YORK", "PENNSYLVANIA")] <- "Northeast"
Dataset_new$Region[Dataset_new$STATE%in% c("ILLINOIS", "INDIANA", "MICHIGAN", "OHIO", "WISCONSIN", "IOWA", "KANSAS", "MINNESOTA", "MISSOURI", "NEBRASKA","NORTH DAKOTA", "SOUTH DAKOTA")] <- "Midwest"
Dataset_new$Region[Dataset_new$STATE %in% c("DELAWARE", "FLORIDA",
"GEORGIA", "MARYLAND",
"NORTH CAROLINA", "SOUTH CAROLINA",
"VIRGINIA", "DISTRICT OF COLUMBIA",
"WEST VIRGINIA", "ALABAMA", "KENTUCKY",
"MISSISSIPPI", "TENNESSEE", "ARKANSAS",
"LOUISIANA", "OKLAHOMA", "TEXAS"
)] <- "South"
Dataset_new$Region[Dataset_new$STATE %in% c("ARIZONA", "COLORADO", "IDAHO",
"MONTANA", "NEVADA", "NEW MEXICO",
"UTAH", "WYOMING", "ALASKA",
"CALIFORNIA", "HAWAII", "OREGON",
"WASHINGTON"
)] <- "West"
south_data <-Dataset_new[Dataset_new$Region == 'South',]
north_data <-Dataset_new[Dataset_new$Region == 'Northeast',]
midwest_data <-Dataset_new[Dataset_new$Region == 'Midwest',]
west_data <-Dataset_new[Dataset_new$Region == 'West',]
# Intensity plotting for events vs state-- each point is scaled according to number of occurence
# South Region-
south_table <- table(midwest_data)
table_s<- as.data.frame(south_table)
names(table_s)<- c("State","Event","Region","Frequency")
summary(table_s$Frequency)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.00 0.00 6.00 49.84 46.00 1034.00p<- subset(table_s,subset=Frequency>50, select= c(State,Event,Frequency))
qplot(p$State, p$Event, data=p, size=Frequency,col= Event, alpha=I(6/10), main="Intensity of Events in Mid-West Region",
xlab="State", ylab="Events Type")+theme(axis.text.x = element_text(angle = 90, hjust =1))
Here We can see that number or frequency of occurrence of event types are represented by the size of the circle as shown in Frequency. This graph gives the information on number of events occurring in a state. We can see that Thunderstorm and Hail are occurring in every state. Comparison is possible of any particular event types between state. Similar graphs are drawn for other regions as well.
# North Region-
north_table <- table(north_data)
table_n<- as.data.frame(north_table)
names(table_n)<- c("State","Event","Region","Frequency")
summary(table_n$Frequency)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.0 0.0 1.0 17.9 13.5 486.0p2<- subset(table_n,subset=Frequency>10, select= c(State,Event,Frequency))
qplot(p2$State, p2$Event, data=p2, size=Frequency,col= Event, alpha=I(6/10), main="Intensity of Events in North Region",
xlab="State", ylab="Events Type")+theme(axis.text.x = element_text(angle = 90, hjust =1))
# MidWest Region-
midwest_table <- table(midwest_data)
table_mw<- as.data.frame(midwest_table)
names(table_mw)<- c("State","Event","Region","Frequency")
summary(table_mw$Frequency)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.00 0.00 6.00 49.84 46.00 1034.00p1<- subset(table_mw,subset=Frequency>50, select= c(State,Event,Frequency))
qplot(p1$State, p1$Event, data=p1, size=Frequency,col= Event, alpha=I(6/10), main="Intensity of Events in Mid-West Region",
xlab="State", ylab="Events Type")+theme(axis.text.x = element_text(angle = 45, hjust =1))
# West Region-
west_data <- na.omit(west_data)
west_table <- table(west_data)
table_w<- as.data.frame(west_table)
names(table_w)<- c("State","Event","Region","Frequency")
summary(table_w$Frequency)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.00 0.00 1.00 19.85 10.00 756.00p4<- subset(table_w,subset=Frequency>0, select= c(State,Event,Frequency))
qplot(p4$State, p4$Event, data=p4, size=Frequency,col= Event, alpha=I(6/10), main="Intensity of Events in West Region",
xlab="State", ylab="Events Type")+theme(axis.text.x = element_text(angle = 90, hjust =1))
Below are the plots in US map showing the similar distribution as shown above. Here top events are taken :Falsh flood, flood, hail, tornado and thunderstorm. Though Hurricane(Typhoon), Tropical Storm, Storm surge/Tide, wildfire are also topmost damaging events, longitudes and latitudes are missing in data. Thus we haven’t plotted these events in US map.
library("Rcpp")
library("ggmap")
library("qmap")
Dataset1<-dataset2016[,c(8,9,11,13,45,46,47,48)]
Dataset2<- na.omit(Dataset1)
# Making subset according to events, taking top 5 damage causing events
thunder_plot<- subset(Dataset2,EVENT_TYPE=="Thunderstorm Wind")
thunder_plot<- na.omit(thunder_plot)
FFlood_plot<- subset(Dataset2,EVENT_TYPE=="Flash Flood")
FFlood_plot<- na.omit(FFlood_plot)
Flood_plot<- subset(Dataset2,EVENT_TYPE=="Flood")
Flood_plot<- na.omit(Flood_plot)
hail_plot<- subset(Dataset2,EVENT_TYPE=="Hail")
hail_plot<- na.omit(hail_plot)
tornado_plot<- subset(Dataset2,EVENT_TYPE=="Tornado")
tornado_plot<- na.omit(tornado_plot)Here we can see which region is more affected by these events clearly. Thunderstorm is more prevalent and distributed over large area.
usmap<-qmap("United States",zoom = 4)
usmap+geom_point(aes(x=BEGIN_LON,y=BEGIN_LAT),data=FFlood_plot)+ggtitle("Thunderstorm Plot")
usmap+geom_point(aes(x=BEGIN_LON,y=BEGIN_LAT),data=FFlood_plot)+ggtitle("Flash Flood Plot")
usmap+geom_point(aes(x=BEGIN_LON,y=BEGIN_LAT),data=Flood_plot)+ggtitle("Flood Plot")
usmap+geom_point(aes(x=BEGIN_LON,y=BEGIN_LAT),data=hail_plot)+ggtitle("Hail Plot")
usmap+geom_point(aes(x=BEGIN_LON,y=BEGIN_LAT),data=tornado_plot)+ggtitle("Tornado Plot")
2. What is the amunt of damage (property & Crops) caused by each event for year 2016?
For this analysis, we have used the column event types, property damage and crop damage. To focus more on the events which are causing the most damages, we have taken top 10 events instead of considering all the events available in data.
library(dplyr)
library(ggplot2)
library(scales)
library(reshape)
#extract 2016 data
data_2016 = subset (dataall_update, YEAR == 2016)
Dataset4_na_addcol <- data_2016
# adding column TOTAL_DAMAGE
Dataset4_na_addcol$TOTAL_DAMAGE <- (Dataset4_na_addcol$DAMAGE_PROPERTY + Dataset4_na_addcol$DAMAGE_CROPS)
## topten events
Damage <- aggregate(cbind(DAMAGE_PROPERTY, DAMAGE_CROPS,TOTAL_DAMAGE) ~ EVENT_TYPE, data = Dataset4_na_addcol, FUN = sum)
#topten events by property damage
propertyDamage <- Damage[order(Damage$DAMAGE_PROPERTY, decreasing = T), ]
topTen_propertyDamage <- propertyDamage[1:10, ]
topTen_propertyDamage1 <- data.frame(apply(topTen_propertyDamage, 2, unclass))
#topten events by crop damage
cropDamage <- Damage[order(Damage$DAMAGE_CROPS, decreasing = T), ]
topTen_cropDamage <- cropDamage[1:10, ]
topTen_cropDamage1 <- data.frame(apply(topTen_cropDamage, 2, unclass))
#topten events by total(property+ crop) damage
totalDamage <- Damage[order(Damage$TOTAL_DAMAGE, decreasing = T), ]
topTen_totalDamage <- totalDamage[1:10, ]
topTen_totalDamage1 <- data.frame(apply(topTen_totalDamage, 2, unclass))Plots for above data have been created as follows:
black.bold.italic.text <- element_text(face = "bold.italic", color = "black")
#plot for damage property
p <- ggplot(topTen_propertyDamage1, aes(x=EVENT_TYPE, y = DAMAGE_PROPERTY))+geom_bar(stat = "identity", width = 0.1, fill="black")
p+ theme(axis.text.x = element_text(angle = 45, hjust =1),axis.line = element_line(size = 1, colour = "black", linetype = "dashed"),panel.background = element_rect(fill = "white"),panel.grid.major = element_line(colour = "black", linetype = "dotted"), panel.grid.minor = element_line(colour = "black", linetype = "dotted"),
title = black.bold.italic.text, axis.title = black.bold.italic.text)+
labs(title = "Damage of property by events",
x="Event Types", y="Amount of Damage $ (in K)")
Here FlashFlood is the most devastating for property in 2016
#plot for damage crop
p1 <- ggplot(topTen_cropDamage1, aes(x=EVENT_TYPE, y = DAMAGE_CROPS))+geom_bar(stat = "identity", width = 0.1, fill ="orange")
p1+ theme(axis.text.x = element_text(angle = 45, hjust =1),axis.line = element_line(size = 1, colour = "black", linetype = "dashed"),panel.background = element_rect(fill = "white"),panel.grid.major = element_line(colour = "black", linetype = "dotted"), panel.grid.minor = element_line(colour = "black", linetype = "dotted"),
title = black.bold.italic.text, axis.title = black.bold.italic.text)+
labs(title = "Damage of crops by events",
x="Event Types", y="Amount of Damage $ (in K)") 
Flood is the most devastating for crops
Summarized plot of total damage is shown below
#plot for Total damage
p2 <- ggplot(topTen_totalDamage1, aes(x=EVENT_TYPE, y = TOTAL_DAMAGE))+geom_bar(stat = "identity", width = 0.1)
p2+ theme(axis.text.x = element_text(angle = 45, hjust =1),
title = black.bold.italic.text, axis.title = black.bold.italic.text)+
labs(title = "Total Damage by events",
x="Event Types", y="Amount of Damage $ (in K)")
#plot for overall damage with respective damage in property and crops
p3 <- melt(topTen_totalDamage[, c("EVENT_TYPE", "TOTAL_DAMAGE", "DAMAGE_PROPERTY", "DAMAGE_CROPS")], id.vars = 1)
ggplot(p3, aes(x = reorder(EVENT_TYPE, -value), y = value)) + geom_bar(stat = "identity", aes(fill = variable), position = "dodge")+scale_y_continuous(labels = comma)+ theme(axis.text.x = element_text(angle = 45, hjust = 1), axis.line = element_line(size = 1, colour = "black", linetype = "dashed"),
panel.background = element_rect(fill = "white"), panel.grid.major = element_line(colour = "black", linetype = "dotted"),
panel.grid.minor = element_line(colour = "black", linetype = "dotted")) + xlab("Event Type") + ylab("Total amount $ (in K)") + ggtitle("Top ten harmful weather event types in 2016") +
scale_fill_manual(name = "Damage Type", values = c("forestgreen", "skyblue", "orange"),
labels = c("Total Damage", "Property Damage", "Crop Damage"))
From the above graph we can see that event types which are most devastating has more effects on property than crop when compared on the basis of amount.
3. What is trend of top 5 events occurred in 2016 troughout the duration of 2011 -2016
We have taken top 5 events on the basis of total damage on property and crop. We are trying to look into the trend of these top eventsfrom 2011 - 2016. We have top 5 events : FlashFlood, Flood, Hail, Hurricane, Tropical Storm. Along with these, thunderstorm is also taken since it is one of the most happening uniform event in 2016.
#Add column which contains the sum of amount of damage property and damage crops
dataall_na_addcol <- dataall_update[,c(8,9,11,12,13,18,25,26)]
dataall_na_addcol$DAMAGE_OVERALL <- (dataall_na_addcol$DAMAGE_PROPERTY + dataall_na_addcol$DAMAGE_CROPS)
# plotting
dataall_na_addcol$date <- as.POSIXct(dataall_na_addcol$BEGIN_DATE_TIME,format = "%d-%b-%y %H:%M:%S")
# making subset of the damage as thuderstorm is the most occuring event each year-
dataset_thunder <- subset(dataall_na_addcol,EVENT_TYPE=="Thunderstorm Wind")
dataset_thunder <- dataset_thunder[c(9,10)]
dataset_thunder$date <- as.Date(dataset_thunder$date)
# for relative scaling, removed data with damage 0 and greater than 50000
dataset_thunder1 <- subset(dataset_thunder,subset = DAMAGE_OVERALL>0 & DAMAGE_OVERALL<50000 , select = c(DAMAGE_OVERALL,date))
# plot time series
ts_thunder<- ts(dataset_thunder1$DAMAGE_OVERALL,start=c(2011,1),end=c(2016,12),frequency = 12)
plot(ts_thunder, type="l",col="blue", xlab = "Date",ylab="Relative Damage", main = "Time series of Thunderstorm")
We can see that thunderstorm is almost occurring in every year through out the months. Here 0 is relative rather than the absolute value.
Similar trend plots are done for top 5 events as below
# Creating plot for top 5 damage causing events- 1. Flash Flood
dataset_flashflood <- subset(dataall_na_addcol,EVENT_TYPE=="Flash Flood")
dataset_flashflood <- dataset_flashflood[c(9,10)]
dataset_flashflood$date <- as.Date(dataset_flashflood$date)
dataset_flashflood1 <- subset(dataset_flashflood,subset = DAMAGE_OVERALL>0 , select = c(DAMAGE_OVERALL,date))
ts_flashflood<- ts(dataset_flashflood1$DAMAGE_OVERALL,start=c(2011,1),end=c(2016,12),frequency = 12)
plot(ts_flashflood, type="l",col="blue", xlab = "Date",ylab="Relative Damage", main = "Time series of Flash Flood")
# 2. Flood
dataset_flood <- subset(dataall_na_addcol,EVENT_TYPE=="Flood")
dataset_flood <- dataset_flood[c(9,10)]
dataset_flood$date <- as.Date(dataset_flood$date)
dataset_flood1 <- subset(dataset_flood,subset = DAMAGE_OVERALL>0 , select = c(DAMAGE_OVERALL,date))
ts_flood<- ts(dataset_flood1$DAMAGE_OVERALL,start=c(2011,1),end=c(2016,12),frequency = 12)
plot(ts_flood, type="l",col="blue", xlab = "Date",ylab="Relative Damage", main = "Time series of Flood")
# 3. Hail
dataset_hail <- subset(dataall_na_addcol,EVENT_TYPE=="Hail")
dataset_hail <- dataset_hail[c(9,10)]
dataset_hail$date <- as.Date(dataset_hail$date)
dataset_hail1 <- subset(dataset_hail,subset = DAMAGE_OVERALL>0 , select = c(DAMAGE_OVERALL,date))
ts_hail<- ts(dataset_hail1$DAMAGE_OVERALL,start=c(2011,1),end=c(2016,12),frequency = 12)
plot(ts_hail, type="l",col="blue", xlab = "Date",ylab="Relative Damage", main = "Time series of Hail")
# 4. Hurricane
dataset_hurricane <- subset(dataall_na_addcol,EVENT_TYPE=="Hurricane")
dataset_hurricane <- dataset_hurricane[c(9,10)]
dataset_hail$date <- as.Date(dataset_hail$date)
dataset_hurricane1 <- subset(dataset_hurricane,subset = DAMAGE_OVERALL>0 , select = c(DAMAGE_OVERALL,date))
ts_hurricane<- ts(dataset_hurricane1$DAMAGE_OVERALL,start=c(2011,1),end=c(2016,12),frequency = 12)
plot(ts_hurricane, type="l",col="blue", xlab = "Date",ylab="Relative Damage", main = "Time series of Hurricane")
# 5. Tropical Storm
dataset_TropicalStorm <- subset(dataall_na_addcol,EVENT_TYPE=="Tropical Storm")
dataset_TropicalStorm <- dataset_TropicalStorm[c(9,10)]
dataset_TropicalStorm$date <- as.Date(dataset_TropicalStorm$date)
dataset_TropicalStorm1 <- subset(dataset_TropicalStorm,subset = DAMAGE_OVERALL>0 , select = c(DAMAGE_OVERALL,date))
ts_TropicalStorm<- ts(dataset_TropicalStorm1$DAMAGE_OVERALL,start=c(2011,1),end=c(2016,12),frequency = 12)
plot(ts_TropicalStorm, type="l", col="blue", xlab = "Date",ylab="Relative Damage", main = "Time series of Tropical storm")
4.What is the frequency of event occurrence on the monthly basis (month wise occurrence) for year 2016?
library(knitr)
library(ggplot2)
require(gridExtra)
Dataset=read.csv('StormEvents_details-ftp_v1.0_d2016_c20170918.csv')
Dataset3_month <- Dataset[,c(12,13)]
# Create event table.
event_table=table(Dataset3_month$EVENT_TYPE)
event_table_df<- as.data.frame(event_table)
names(event_table_df)<- c("Event","Frequency")
ggplot(event_table_df, aes(x = Frequency)) + geom_density() +scale_y_continuous(labels = comma) # this plot is to find out which event should be considered for plotting in clustered graph
# So event above 5000 frequency should be considered while plotting
# Making subset of the graphs with frequency above the median of the EVenet Frequency so that we only consider the higher occuring events
event_table_df<- subset(event_table_df,subset=Frequency>median(event_table_df$Frequency), select= c(Event))
#Showing the month with frequency of event occurance-
month_table <- table(Dataset3_month)
table_M<- as.data.frame(month_table)
names(table_M)<- c("Month","Event","Frequency")
mean(table_M$Frequency)## [1] 93.265summary(table_M)## Month Event Frequency
## April : 50 Astronomical Low Tide: 12 Min. : 0.00
## August : 50 Avalanche : 12 1st Qu.: 0.00
## December: 50 Blizzard : 12 Median : 3.00
## February: 50 Coastal Flood : 12 Mean : 93.27
## January : 50 Cold/Wind Chill : 12 3rd Qu.: 48.00
## July : 50 Debris Flow : 12 Max. :5463.00
## (Other) :300 (Other) :528highfreq_event<- subset(table_M,subset=Frequency>350, select= c(Month,Event,Frequency))
highfreq_event1p<- subset(table_M,subset=Frequency>48, select= c(Month,Event,Frequency))
q1<- c("January", "February","March")
q2<- c("April", "May","June")
q3<- c("July", "August","September")
q4<- c("October","November", "December")
highfreq_eventq1<- subset(highfreq_event,subset=Month %in% q1, select= c(Month,Event,Frequency))
highfreq_eventq2<- subset(highfreq_event,subset=Month %in% q2, select= c(Month,Event,Frequency))
highfreq_eventq3<- subset(highfreq_event,subset=Month %in% q3, select= c(Month,Event,Frequency))
highfreq_eventq4<- subset(highfreq_event,subset=Month %in% q4, select= c(Month,Event,Frequency))
#Creating Subset of data for each month-
jan_data <-Dataset3_month[Dataset3_month$MONTH_NAME == 'January',]
feb_data <-Dataset3_month[Dataset3_month$MONTH_NAME == 'February',]
mar_data <-Dataset3_month[Dataset3_month$MONTH_NAME == 'March',]
apr_data <-Dataset3_month[Dataset3_month$MONTH_NAME == 'April',]
may_data <-Dataset3_month[Dataset3_month$MONTH_NAME == 'May',]
jun_data <-Dataset3_month[Dataset3_month$MONTH_NAME == 'June',]
jul_data <-Dataset3_month[Dataset3_month$MONTH_NAME == 'July',]
aug_data <-Dataset3_month[Dataset3_month$MONTH_NAME == 'August',]
sep_data <-Dataset3_month[Dataset3_month$MONTH_NAME == 'September',]
oct_data <-Dataset3_month[Dataset3_month$MONTH_NAME == 'October',]
nov_data <-Dataset3_month[Dataset3_month$MONTH_NAME == 'November',]
dec_data <-Dataset3_month[Dataset3_month$MONTH_NAME == 'December',]
# Creating Plot For each month-
table_16 <- table(jan_data$EVENT_TYPE)
df16<- as.data.frame(table_16)
names(df16)<- c("EventType","Frequency")
summary(df16)# Check for Q3 So as to select the event greater than third Quartile## EventType Frequency
## Astronomical Low Tide: 1 Min. : 0.00
## Avalanche : 1 1st Qu.: 0.00
## Blizzard : 1 Median : 7.50
## Coastal Flood : 1 Mean : 61.52
## Cold/Wind Chill : 1 3rd Qu.: 51.25
## Debris Flow : 1 Max. :754.00
## (Other) :44df16.1<- subset(df16,subset=Frequency>50, select= c(EventType,Frequency))
pjan <- ggplot(df16.1, aes(x=EventType, y = Frequency))+geom_col()+ theme(axis.text.x = element_text(angle = 45, hjust =1))+labs(title = "January Events",x="Event", y="Frequency")
table_feb <- table(feb_data$EVENT_TYPE)
data_feb<- as.data.frame(table_feb)
names(data_feb)<- c("EventType","Frequency")
summary(data_feb)# Check for Q3 So as to select the event greater than third Quartile## EventType Frequency
## Astronomical Low Tide: 1 Min. : 0.00
## Avalanche : 1 1st Qu.: 0.00
## Blizzard : 1 Median : 8.50
## Coastal Flood : 1 Mean : 97.48
## Cold/Wind Chill : 1 3rd Qu.: 84.50
## Debris Flow : 1 Max. :814.00
## (Other) :44data_feb<- subset(data_feb,subset=Frequency>51, select= c(EventType,Frequency))
pfeb <- ggplot(data_feb, aes(x=EventType, y = Frequency))+geom_col()+ theme(axis.text.x = element_text(angle = 45, hjust =1))+labs(title = "February Events",x="Event", y="Frequency")
table_mar <- table(mar_data$EVENT_TYPE)
data_mar<- as.data.frame(table_mar)
names(data_mar)<- c("EventType","Frequency")
summary(data_mar)# Check for Q3 So as to select the event greater than third Quartile## EventType Frequency
## Astronomical Low Tide: 1 Min. : 0.00
## Avalanche : 1 1st Qu.: 0.00
## Blizzard : 1 Median : 3.00
## Coastal Flood : 1 Mean : 78.22
## Cold/Wind Chill : 1 3rd Qu.: 37.75
## Debris Flow : 1 Max. :918.00
## (Other) :44data_mar<- subset(data_mar,subset=Frequency>38, select= c(EventType,Frequency))
pmar <- ggplot(data_mar, aes(x=EventType, y = Frequency))+geom_col()+ theme(axis.text.x = element_text(angle = 45, hjust =1))+labs(title = "March Events",x="Event", y="Frequency")
table_apr <- table(apr_data$EVENT_TYPE)
data_apr<- as.data.frame(table_apr)
names(data_apr)<- c("EventType","Frequency")
summary(data_apr)# Check for Q3 So as to select the event greater than third Quartile## EventType Frequency
## Astronomical Low Tide: 1 Min. : 0.00
## Avalanche : 1 1st Qu.: 0.00
## Blizzard : 1 Median : 3.00
## Coastal Flood : 1 Mean : 98.92
## Cold/Wind Chill : 1 3rd Qu.: 54.75
## Debris Flow : 1 Max. :1386.00
## (Other) :44data_apr<- subset(data_apr,subset=Frequency>54, select= c(EventType,Frequency))
papr <- ggplot(data_apr, aes(x=EventType, y = Frequency))+geom_col()+ theme(axis.text.x = element_text(angle = 45, hjust =1))+labs(title = "April Events",x="Event", y="Frequency")
grid.arrange(pjan, pfeb, ncol = 2)
grid.arrange(pmar, papr, ncol = 2)
In above graph we can see the frequency of events and which event is occurring most in the particular months. Plots are created for other months till december.
table_may <- table(may_data$EVENT_TYPE)
data_may<- as.data.frame(table_may)
names(data_may)<- c("EventType","Frequency")
summary(data_may)# Check for Q3 So as to select the event greater than third Quartile## EventType Frequency
## Astronomical Low Tide: 1 Min. : 0.00
## Avalanche : 1 1st Qu.: 0.00
## Blizzard : 1 Median : 3.00
## Coastal Flood : 1 Mean : 112.10
## Cold/Wind Chill : 1 3rd Qu.: 18.25
## Debris Flow : 1 Max. :2576.00
## (Other) :44data_may<- subset(data_may,subset=Frequency>18, select= c(EventType,Frequency))
pmay <- ggplot(data_may, aes(x=EventType, y = Frequency))+geom_col()+ theme(axis.text.x = element_text(angle = 45, hjust =1))+labs(title = "May Events",x="Event", y="Frequency")
table_jun<- table(jun_data$EVENT_TYPE)
data_jun<- as.data.frame(table_jun)
names(data_jun)<- c("EventType","Frequency")
summary(data_jun)# Check for Q3 So as to select the event greater than third Quartile## EventType Frequency
## Astronomical Low Tide: 1 Min. : 0.00
## Avalanche : 1 1st Qu.: 0.00
## Blizzard : 1 Median : 3.00
## Coastal Flood : 1 Mean : 132.24
## Cold/Wind Chill : 1 3rd Qu.: 31.75
## Debris Flow : 1 Max. :3461.00
## (Other) :44data_jun<- subset(data_jun,subset=Frequency>32, select= c(EventType,Frequency))
pjun <- ggplot(data_jun, aes(x=EventType, y = Frequency))+geom_col()+ theme(axis.text.x = element_text(angle = 45, hjust =1))+labs(title = "June Events",x="Event", y="Frequency")
grid.arrange(pmay, pjun, ncol = 2)
table_jul<- table(jul_data$EVENT_TYPE)
data_jul<- as.data.frame(table_jul)
names(data_jul)<- c("EventType","Frequency")
summary(data_jul)# Check for Q3 So as to select the event greater than third Quartile## EventType Frequency
## Astronomical Low Tide: 1 Min. : 0.0
## Avalanche : 1 1st Qu.: 0.0
## Blizzard : 1 Median : 0.0
## Coastal Flood : 1 Mean : 197.9
## Cold/Wind Chill : 1 3rd Qu.: 53.5
## Debris Flow : 1 Max. :5463.0
## (Other) :44data_jul<- subset(data_jul,subset=Frequency>53, select= c(EventType,Frequency))
pjul <- ggplot(data_jul, aes(x=EventType, y = Frequency))+geom_col()+ theme(axis.text.x = element_text(angle = 45, hjust =1))+labs(title = "July Events",x="Event", y="Frequency")
table_aug<- table(aug_data$EVENT_TYPE)
data_aug<- as.data.frame(table_aug)
names(data_aug)<- c("EventType","Frequency")
summary(data_aug)# Check for Q3 So as to select the event greater than third Quartile## EventType Frequency
## Astronomical Low Tide: 1 Min. : 0.00
## Avalanche : 1 1st Qu.: 0.00
## Blizzard : 1 Median : 0.50
## Coastal Flood : 1 Mean : 96.64
## Cold/Wind Chill : 1 3rd Qu.: 39.00
## Debris Flow : 1 Max. :1710.00
## (Other) :44data_aug<- subset(data_aug,subset=Frequency>39, select= c(EventType,Frequency))
paug <- ggplot(data_aug, aes(x=EventType, y = Frequency))+geom_col()+ theme(axis.text.x = element_text(angle = 45, hjust =1))+labs(title = "August Events",x="Event", y="Frequency")
grid.arrange(pjul, paug, ncol = 2)
table_sep<- table(sep_data$EVENT_TYPE)
data_sep<- as.data.frame(table_sep)
names(data_sep)<- c("EventType","Frequency")
summary(data_sep)# Check for Q3 So as to select the event greater than third Quartile## EventType Frequency
## Astronomical Low Tide: 1 Min. : 0.00
## Avalanche : 1 1st Qu.: 0.00
## Blizzard : 1 Median : 5.00
## Coastal Flood : 1 Mean : 65.66
## Cold/Wind Chill : 1 3rd Qu.: 26.25
## Debris Flow : 1 Max. :747.00
## (Other) :44data_sep<- subset(data_sep,subset=Frequency>27, select= c(EventType,Frequency))
psep <- ggplot(data_sep, aes(x=EventType, y = Frequency))+geom_col()+ theme(axis.text.x = element_text(angle = 45, hjust =1))+labs(title = "September Events",x="Event", y="Frequency")
table_oct<- table(oct_data$EVENT_TYPE)
data_oct<- as.data.frame(table_oct)
names(data_oct)<- c("EventType","Frequency")
summary(data_oct)# Check for Q3 So as to select the event greater than third Quartile## EventType Frequency
## Astronomical Low Tide: 1 Min. : 0.00
## Avalanche : 1 1st Qu.: 0.00
## Blizzard : 1 Median : 7.50
## Coastal Flood : 1 Mean : 46.88
## Cold/Wind Chill : 1 3rd Qu.: 49.00
## Debris Flow : 1 Max. :331.00
## (Other) :44data_oct<- subset(data_oct,subset=Frequency>49, select= c(EventType,Frequency))
poct <- ggplot(data_oct, aes(x=EventType, y = Frequency))+geom_col()+ theme(axis.text.x = element_text(angle = 45, hjust =1))+labs(title = "October Events",x="Event", y="Frequency")
grid.arrange(psep, poct, ncol = 2)
table_nov<- table(nov_data$EVENT_TYPE)
data_nov<- as.data.frame(table_nov)
names(data_nov)<- c("EventType","Frequency")
summary(data_nov)# Check for Q3 So as to select the event greater than third Quartile## EventType Frequency
## Astronomical Low Tide: 1 Min. : 0.00
## Avalanche : 1 1st Qu.: 0.00
## Blizzard : 1 Median : 1.00
## Coastal Flood : 1 Mean : 41.32
## Cold/Wind Chill : 1 3rd Qu.: 39.25
## Debris Flow : 1 Max. :610.00
## (Other) :44data_nov<- subset(data_nov,subset=Frequency>39, select= c(EventType,Frequency))
pnov <- ggplot(data_nov, aes(x=EventType, y = Frequency))+geom_col()+ theme(axis.text.x = element_text(angle = 45, hjust =1))+labs(title = "November Events",x="Event", y="Frequency")
table_dec<- table(dec_data$EVENT_TYPE)
data_dec<- as.data.frame(table_dec)
names(data_dec)<- c("EventType","Frequency")
summary(data_dec)# Check for Q3 So as to select the event greater than third Quartile## EventType Frequency
## Astronomical Low Tide: 1 Min. : 0.00
## Avalanche : 1 1st Qu.: 0.00
## Blizzard : 1 Median : 6.00
## Coastal Flood : 1 Mean : 90.28
## Cold/Wind Chill : 1 3rd Qu.: 71.50
## Debris Flow : 1 Max. :738.00
## (Other) :44data_dec<- subset(data_dec,subset=Frequency>71, select= c(EventType,Frequency))
pdec <- ggplot(data_dec, aes(x=EventType, y = Frequency))+geom_col()+ theme(axis.text.x = element_text(angle = 45, hjust =1))+labs(title = "December Events",x="Event", y="Frequency")
grid.arrange(pnov, pdec, ncol = 2)
Following graph shows the occurrence of total events in each month. We can see that July has the highest number of events occurring followed by June and May. This graph helps in understanding of event counts against month of 2016.
ggplot(Dataset3_month, aes(x = MONTH_NAME)) + geom_bar()+theme(axis.text.x = element_text(angle = 90, hjust =1))+labs(title = "Number of Events Per Month",x="Month", y="Event Count")
Following graph shows the nnumber of events in each month with the distribution/frequency of all events. Here, in July (month having the most events) has thunderstorm and hail as predominant events. Throughout the months, these two events are mostly occurring.
# THIS SHOWS WHICH MONTH HAS HOW MANY EVENTS OCCURING
ggplot(Dataset3_month, aes(x = MONTH_NAME, fill = EVENT_TYPE)) + geom_bar()+
theme(axis.text.x = element_text(angle = 90, hjust =1))+
labs(title = "Clustered Graph Showing Month and all Events",x="Month", y="Event Count")
Following graph is same as the above one, only difference is only the high occurring event types are taken on the basis of the analysis above. Here higher frequency greater than 350 is taken. Thus in october, not any event has occured more than 350 times. Thus its not seen in graph.
ggplot(highfreq_event, aes(x = Month, fill = Event)) + geom_bar()+
theme(axis.text.x = element_text(angle = 90, hjust =1))+labs(title = "Clustered Graph Showing Month and high occuringEvents",x="Month", y="Event Count")
In the graph below distribution of months is shown per event. We can see clearly thunderstorm is occurring in each month, followed by hail,flood.
ggplot(highfreq_event, aes(x = Month, fill = Event)) + geom_bar()+
theme(axis.text.x = element_text(angle = 90, hjust =1))+labs(title = "Clustered Graph Showing Month and high occuringEvents",x="Month", y="Event Count")
Following graphs are refined version of graphs made earlier. Events with higher frequency of occurrence are only shown, thus less number of events can be seen thatn in previous graphs, and will be easy to analyze.
## Graphing for higher occuring events-
sub_event<- event_table_df[["Event"]]
#making subset using event_table_df
highevents<- subset(Dataset3_month, EVENT_TYPE %in% sub_event )
# Clustered Graphs
ggplot(highevents, aes(x = MONTH_NAME, fill = EVENT_TYPE)) + geom_bar()+theme(axis.text.x = element_text(angle = 90, hjust =1))+labs(title = "Cluster of Events per Month",x="Month", y="Event Count")
ggplot(highevents, aes(x = EVENT_TYPE, fill = MONTH_NAME)) + geom_bar()+ theme(axis.text.x = element_text(angle = 90, hjust =1))+labs(title = "Cluster of months per event type",x="Events", y="Month")
Following 4 graphs are shown to validate the analysis and graphs above. Here if we focus n thunderstorm and hail, they are occurring in each month while other events are discrete. Here higher frequency greater than 350 is taken. Thus in october, not any event has occured more than 350 times. Thus its not seen in graph.
# Line Graph Showing the frequency of Events by month (Showing Seasonality of Event Types)
ggplot(data=highfreq_event, aes(x=Month, y=Frequency, group=Event)) +
geom_line(aes(color=Event))+
geom_point(aes(color=Event))+theme(axis.text.x = element_text(angle = 90, hjust =1))+labs(title = "Seasonality of Events",x="Month", y="Frequency")
# for quater 1-
ggplotq1<-ggplot(data=highfreq_eventq1, aes(x=Month, y=Frequency, group=Event)) +
geom_line(aes(color=Event))+
geom_point(aes(color=Event))+theme(axis.text.x = element_text(angle = 90, hjust =1))+labs(title = "Events in Quater 1",x="Month", y="Frequency")
# Plot for quater 2-
ggplotq2 <- ggplot(data=highfreq_eventq2, aes(x=Month, y=Frequency, group=Event)) +
geom_line(aes(color=Event))+
geom_point(aes(color=Event))+theme(axis.text.x = element_text(angle = 90, hjust =1))+labs(title = "Events in Quater 2",x="Month", y="Frequency")
grid.arrange(ggplotq1, ggplotq2, ncol = 2)
# Plot for Quater 3-
ggplotq3 <- ggplot(data=highfreq_eventq3, aes(x=Month, y=Frequency, group=Event)) +
geom_line(aes(color=Event))+
geom_point(aes(color=Event))+theme(axis.text.x = element_text(angle = 90, hjust =1))+labs(title = "Events in Quater 3",x="Month", y="Frequency")
# Plot for Quater 4-
ggplotq4 <- ggplot(data=highfreq_eventq4, aes(x=Month, y=Frequency, group=Event)) +
geom_line(aes(color=Event))+
geom_point(aes(color=Event))+theme(axis.text.x = element_text(angle = 90, hjust =1))+labs(title = "Events in Quater 4",x="Month", y="Frequency")
grid.arrange(ggplotq3, ggplotq4, ncol = 2)
5.What is the total number of casualties (injuries and deaths ) in 2016 by each events?
Here we have considered top 10 events on the basis of injuries and deaths and eventaully casualties. Injuries is considered as sum of direct and indirect injuries in data and deaths is considered combining direct and indirect death columns of data. Casualties is calculated adding up both injuries and casualties.
Dataset4_na_addcol$TOTAL_INJURIES <- (Dataset4_na_addcol$INJURIES_DIRECT + Dataset4_na_addcol$INJURIES_INDIRECT)
Dataset4_na_addcol$TOTAL_DEATHS <- (Dataset4_na_addcol$DEATHS_DIRECT + Dataset4_na_addcol$DEATHS_INDIRECT)
Dataset4_na_addcol$TOTAL_CASUALTIES <- (Dataset4_na_addcol$TOTAL_INJURIES + Dataset4_na_addcol$TOTAL_DEATHS)
# topten events for each case
Injuries <- aggregate(cbind(INJURIES_DIRECT,INJURIES_INDIRECT,TOTAL_INJURIES) ~ EVENT_TYPE, data = Dataset4_na_addcol, FUN = sum )
Deaths <- aggregate(cbind(DEATHS_DIRECT,DEATHS_INDIRECT,TOTAL_DEATHS) ~ EVENT_TYPE, data = Dataset4_na_addcol, FUN = sum)
Casualties <- aggregate(cbind(TOTAL_INJURIES,TOTAL_DEATHS,TOTAL_CASUALTIES) ~ EVENT_TYPE, data = Dataset4_na_addcol, FUN = sum)
totalInjuries <- Injuries[order(Injuries$TOTAL_INJURIES, decreasing = T), ]
topTen_totalInjuries <- totalInjuries[1:10, ]
totalDeaths <- Deaths[order(Deaths$TOTAL_DEATHS, decreasing = T), ]
topTen_totalDeaths <- totalDeaths[1:10, ]
totalCasualties <- Casualties[order(Casualties$TOTAL_CASUALTIES, decreasing = T), ]
topTen_totalCasualties <- totalCasualties[1:10, ]
# Plots
#plot for injuries
p4 <- ggplot(topTen_totalInjuries, aes(x=EVENT_TYPE, y = TOTAL_INJURIES))+geom_bar(stat = "identity", width = 0.1, fill = "blue")
p4+ theme(axis.text.x = element_text(angle = 45, hjust =1),
title = black.bold.italic.text, axis.title = black.bold.italic.text)+
labs(title = "Total injuries by events",
x="Event Types", y="Total Number of Injuries")
In above graph, Torando is the most devastating event causing injuries in 2016
#plot for Deaths
p5 <- ggplot(topTen_totalDeaths, aes(x=EVENT_TYPE, y = TOTAL_DEATHS))+geom_bar(stat = "identity", width = 0.1, fill = "forestgreen")
p5+ theme(axis.text.x = element_text(angle = 45, hjust =1),
title = black.bold.italic.text, axis.title = black.bold.italic.text)+
labs(title = "Total deaths by events",
x="Event Types", y="Total Number of Deaths")
In the above graph, Flash flood and flood are causing most deaths in 2016
#plot for Total Casualties
p6 <- melt(topTen_totalCasualties[, c("EVENT_TYPE", "TOTAL_CASUALTIES", "TOTAL_INJURIES", "TOTAL_DEATHS")], id.vars = 1)
ggplot(p6, aes(x = reorder(EVENT_TYPE, -value), y = value)) + geom_bar(stat = "identity", aes(fill = variable), position = "dodge")+scale_y_continuous(labels = comma)+ theme(axis.text.x = element_text(angle = 45, hjust = 1), axis.line = element_line(size = 1, colour = "black", linetype = "dashed"),
panel.background = element_rect(fill = "white"), panel.grid.major = element_line(colour = "black", linetype = "dotted"),
panel.grid.minor = element_line(colour = "black", linetype = "dotted")) + xlab("Event Type") + ylab("Total Number of Damage") + ggtitle("Top ten harmful weather event types in 2016") +
scale_fill_manual(name = "Damage Type", values = c("forestgreen", "skyblue", "orange"),
labels = c("Total casualties", "injuries", " deaths"))
Above graph shows the total casualties considering deaths and injuries above. We can see that some event like tornado has more injuries than deaths while events like flash flood, rip current and flood are causing more deaths than injuries.
So the takeaway from these analyses are:
In 2016, Tornado is the most devastating event on the basis of casualties - however injuries way higher than deaths for tornado.
In 2016, Flash Flood is the most devastating event on the basis of property damage.
In 2016, Flood is the major cause for crop damage.
Looking at the trend of topmost events of 2016, thunderstorm and hail are occurring almost each month.
Looking at the trend of topmost events of 2016 for duration 2011-2016, thunderstorm, hurricane are more uniform than other events.